PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of scanLink to Publisher's site
 
Soc Cogn Affect Neurosci. 2015 July; 10(7): 978–986.
Published online 2014 November 17. doi:  10.1093/scan/nsu146
PMCID: PMC4483568

Neural correlates of generation and inhibition of verbal association patterns in mood disorders

Abstract

Objectives: Thought disorders such as rumination or flight of ideas are frequent in patients with mood disorders, and not systematically linked to mood state. These symptoms point to anomalies in cognitive processes mediating the generation and control of thoughts; for example, associative thinking and inhibition. However, their neural substrates are not known. Method: To obtain an ecological measure of neural processes underlying the generation and suppression of spontaneous thoughts, we designed a free word association task during fMRI allowing us to explore verbal associative patterns in patients with mood disorders and matched controls. Participants were presented with emotionally negative, positive or neutral words, and asked to produce two words either related or unrelated to these stimuli. Results: Relative to controls, patients produced a reverse pattern of answer typicality for the related vs unrelated conditions. Controls activated larger semantic and executive control networks, as well as basal ganglia, precuneus and middle frontal gyrus. Unlike controls, patients activated fusiform gyrus, parahippocampal gyrus and medial prefrontal cortex for emotional stimuli. Conclusions: Mood disorder patients are impaired in automated associative processes, but prone to produce more unique/personal associations through activation of memory and self-related areas.

Keywords: thoughts, free word association, inhibition, mood disorders, fMRI

INTRODUCTION

Thought processes are often impaired in mood disorders, ranging from racing thoughts to ruminations and thought inhibition (Goodwin and Jamison, 2007). This phenomenology might reflect an impairment in more basic cognitive functions leading to abnormal patterns of associations in the flow of mental processes that generate spontaneous thoughts (Bar et al., 2007; Bar, 2009). Both positive and negative mood states have often been related to distinct associative patterns in cognitive operations, broad or narrow, respectively (Fredrickson, 2001; Pronin and Jacobs, 2008; Mason and Bar, 2012). However, these correlations are not absolute because, in some cases, low mood is accompanied with subjective complaints of numerous and racing thoughts (Akiskal and Benazzi, 2004; Piguet et al., 2010; Desseilles et al., 2012; Keizer et al., 2013).

Given these observations, we postulated that patterns of thought associations might be disrupted in mood disorders. In a recent model of thought disorders (Piguet et al., 2010), we proposed that distinct cognitive abilities underlying mental flexibility and inhibition may constitute core processes whose dysfunction may underlie the phenomenology of thought disorders. Specifically, crowded or racing thoughts might stem from an imbalance between mechanisms responsible for the production (on the one hand) and suppression (on the other hand) of associative thoughts. Anomalies in these processes may represent a common deficit in the spectrum of mood disorders, shared across different diagnostic subcategories (e.g. Major Depressive Disorder [MDD] or Bipolar Disorder [BD]) (Marvel and Paradiso, 2004). Here, we therefore aimed at investigating these traits in a dimensional approach framework (Narrow and Kuhl, 2011; Vieta and Valentí, 2013), across a group of patients suffering from mood disorders with various clinical presentations.

In order to test both production and inhibition components of associative thought processes in the same protocol, we designed a free word association task with two conditions. The first ‘close’ condition asked participants to generate spontaneous/intuitive associations with a word cue, whereas the second ‘remote’ condition asked them to produce responses unrelated to the word cue. We used positive, negative and neutral word cues in order to also examine whether emotional meaning had an influence on activated representations and associations. This task thus allowed us to probe for generation and inhibition processes in associative thinking within the same paradigm. To the best of our knowledge, these processes have not previously been investigated in mood disorders.

The ‘close’ condition in our task resembles verbal fluency tests commonly used in neuropsychology, which assess the initiation of word production and strength of semantic associations (Schwartz et al., 2003). However, compared with the traditional cued word fluency (e.g. phonology- or category-based), free word association tasks are closer to natural thought processes. We also labeled this condition as ‘automatic’ based on the theory of automatic spreading of activation in semantic networks (Manschreck et al., 1988; Spitzer, 1997). Thus, the production of close associates was postulated to reflect a ‘pure’ diffusion of activation through conceptual associations. The ‘remote’ condition in our paradigm was based on the Hayling Sentence Completion Test (Burgess and Shallice, 1996) that requires participants to generate either usual or unusual words in response to a sentence stem, and is used to study both verbal semantic associations and cognitive control. Contrarily to the close condition, producing remote associations requires the retrieval of words that are not linked through direct, automatic associations (e.g. table-giraffe instead of table-chair). This process is effortful and depends on executive control function; therefore, we also labeled this condition as ‘inhibition’ (Noël et al., 2012). This process is mediated by the prefrontal cortex, in particular the right inferior frontal gyrus (Volle et al., 2012), that has been implicated in inhibition of automatisms and prepotent responses across many modalities, motor, verbal and other (e.g. Konishi et al., 1999; Xue et al., 2008). Deficits in this task have been found in bipolar (Drysdale et al., 2013; Wang et al., 2013a) and unipolar patients (Gohier et al., 2009); therefore, we predicted that patients would perform worse than controls for the ‘remote’ condition.

Few neuroimaging studies have used a free word association task, probably because these are less constrained than verbal fluency tasks (where participants have to produce words starting by a specific letter or belonging to a specific semantic category). For example, a study comparing free word association and word generation in healthy volunteers showed that both tasks recruited an overlapping semantic network, predominantly located in fronto-temporal areas of the left hemisphere (Wende et al., 2012), in line with current knowledge on semantic systems (e.g. Binder et al., 2009). Among critical brain regions allowing these flexible binding operations between semantic representations, the hippocampus and parahippocampal cortex (known for their role in autobiographical encoding and retrieval in memory) might also play an important role and appear differentially recruited during free verbal association, in comparison to semantic verbal fluency (Bar et al., 2007; Kircher et al., 2008; Whitney et al., 2009). A few studies used the Hayling task in mood disorders, and found greater activity during sentence completion in limbic areas, such as ventromedial prefrontal cortex and ventral striatum for bipolar patients (McIntosh et al., 2008), or amygdala for their healthy relatives (Whalley et al., 2011). To our knowledge, only one pioneer study has performed this task, including the inhibition condition, in bipolar patients, and has found a negative correlation between error scores and cortical volume of right parietal areas, interpreted as failure of inhibition (Haldane et al., 2008). This condition has been found to differentially activate the precuneus and middle/superior frontal gyrus in healthy participants (HPs) (Collette et al., 2001; Allen et al., 2008). However, little is known about the brain systems underlying the production and inhibition of spontaneous associations in this task, and how these cognitive functions are modified in mood disorders.

Here, by using a free word association task with both close (automatic) and remote (inhibition) conditions in mood disorder patients, we tested for any systematic change in verbal associative patterns that might represent dimensional marker of mood disorders, and for any influence of the emotional meaning of word cues. We predicted that both associative patterns and inhibition processes might be altered in mood disorders, across diagnostic boundaries. We also predicted that fMRI should reveal distinct patterns of activations in semantic network during close and remote conditions, with further differences in patients relative to HPs that should reflect their enhanced self-focus, and/or reduced inhibition abilities.

MATERIALS AND METHODS

Participants

Sixty-four participants speaking French for more than 20 years participated in this study, 32 mood disorder patients and 32 HP matched for age, gender, laterality and level of education. Because behavioral data from one HP could not be analyzed due to recording problems, the final sample included 32 patients and 31 HP. All participants gave informed written consent in accordance with procedures approved by the Ethics Committee of the Geneva University Hospital, in accordance with the Helsinki Declaration of 1975. Participants got modest financial compensation for their participation. Patients were recruited from the department of adult psychiatry of the Geneva University Hospital, as well as through advertising on classified advertisements websites. HPs were also recruited from a local database or through web advertising.

Inclusion criteria for patients were: a diagnostic of MDD or BD, age between 18 and 56 years old, stable medication for at least 4 weeks, with no usual contraindication for MRI. Exclusion criteria for HP were: a past or present history of neurological or psychiatric problem, use of medication and contraindication for MRI. The Mini International Neuropsychiatric Interview (MINI [Sheehan et al., 1998]) and the Structured Clinical Interview for the DSM-IV, Mood Disorders section (SCID-II [First et al., 1997]) were administered to patients and HP (during a separate visit for the patients, maximum 1 week apart). Patients were included regardless of their current mood state. As well as meeting criteria for MDD (N = 9), BD type I (BD-I, N = 7), type II (BD-II, N = 13) or BD-Not Otherwise Specified (N = 3), some patients also met criteria for anxiety disorder (N = 15), borderline personality disorder (N = 10), and Attention Deficit/Hyperactivity Disorder (N = 2), reflecting the high prevalence of comorbid anxiety and personality disorders expected in this population. Laterality was measured by the Edinburgh handedness inventory (Oldfield, 1971). Just before scanning, the current mood state was assessed by the Montgomery–Asberg Depression Rating Scale (MADRS [Montgomery and Asberg, 1979; French version: Pellet et al., 1980]), Young Mania Rating Scale (French version: Young et al., 1978; Favre et al., 2003) through medical interview by trained clinicians. The Hamilton Anxiety Scale (Hamilton, 1959; French version: Pichot et al., 1981) was also rated for both the patients and HP.

Stimuli

Words were selected after a cross-match of (i) existing norms of associations in French in a standard corpus (Ferrand and Alario, 1998; Ferrand, 2001), for both concrete and abstract words, and ii) ratings for emotional valence in the database ‘valemo’ (available on www.lexique.org). We chose French words that had a first word associate representing 30–60% of total answers (medium-size associative sets). We selected a list of 72 words and divided them in three sets of negative, neutral and positive words (see Supplementary Table). Words with a different valence were counterbalanced for gender, frequency in French (movies), length of the word and number of orthographic neighbors. Due to the multiple constraints for carefully counterbalancing this material across conditions, we also added 11 words (three negatives, two neutrals and six positives) coming from the valence database (valemo) that had no norms for associations, but were judged to be similar in other respect (see Supplementary Table, italic type words). Each word appeared only once in each condition (close or remote). The final experimental conditions were: close-positive words, close-negative words, close-neutral words, remote-positive words, remote-negative words and remote-neutral words.

Procedure

The task consisted of two conditions of free semantic association (with three emotion categories each) plus one control condition where participants had to say aloud the grammatical ‘gender’ of the stimulus-word (masculine or feminine). The instruction for the ‘close’ association condition was: ‘Please say aloud the first two words that come to your mind when reading the stimulus’. The instruction for the ‘remote’ association condition was: ‘Please say two words as far as possible of the word presented on the screen, try to avoid any link with the stimulus, even the opposite is still a link’. Participants were also told that they should not repeat the same words, and always try to say two words for each trial. Stimuli for the control condition were taken from the list of neutral stimuli that were not used for the other conditions after counterbalancing.

Imaging data were acquired on TRIO 3T whole-body MR unit (Siemens, Germany) equipped with a 32 channel head coil. A sparse temporal acquisition technique (Hall et al., 1999) was used (with a pause between two MRI images), so that the spoken answers could be recorded with an MRI compatible microphone during an interval free of gradient noise. For more details on the procedure and MRI acquisition parameters, see Figure 1 and Supplementary Material.

Fig. 1
Free Word Association task. Participants were instructed to respond to the word-cue in two different conditions: they had either to ‘say two related words that first come to your mind’ (Close condition) or to ‘say two words as ...

Analysis of behavioral data

Behavioral data analyses were performed using SPSS Statistics 19.0 (SPSS Inc.). For the number of words produced in the free association task, we calculated a global performance index that could account for missing data (two patients and one HP had only two sessions instead of three due to technical recording problems) and computed a percentage of given verbal responses for each recorded condition (number of words produced/number of possible answers). To obtain a qualitative measure of performance (i.e. response originality), we calculated two indices, one based on the frequency of each word, in response to each cue, across all answers from all participants (FREQ), and one based on the number of ‘typical’ association as expected by the normative database (TYP).

FREQ was measured by counting the number of time each word appeared relative to the total number of words produced for one cue in our whole sample, and then computing the mean values (per valence and per condition) for words produced by each subject (see Supplementary Material for an example). Thus, performance was associated with high FREQ values when the subject produced answers similar to the rest of the participants, but low FREQ values when the subject produced more ‘original’ or idiosyncratic answers.

TYP was calculated by computing, for each word cue, how many times the given answer corresponded to one of the five most common ‘expected’ words according to the association database of French words established by Ferrand and Alario (1998). Thus, like for FREQ, the lower the TYP indices were, the more ‘unusual’ or idiosyncratic the subject’s production was (see Supplementary Material for an example).

Analysis of MRI data

During the free association task in the scanner, verbal answers were recorded in a digital format and then transcribed. We modeled the seven experimental conditions described above in the design matrix (close and remote trials, for negative, neutral and positive words + control condition). In addition, the number of words produced for each trial (0,1,2) was computed and used as a parametric modulator in the first level analysis, in order to control for any global variance between conditions and participants unrelated to the associative and cognitive processes of primary interest. Sparse fMRI data were analyzed using standard procedures (see Supplementary Material). Direct statistical group comparisons were obtained using an exclusive masking procedure to reveal voxels that showed increased response in one population relative to the other population. The map constituting the exclusive mask was thresholded at P < 0.05, whereas the contrast to be masked was thresholded at P < 0.001, cluster threshold of more than five voxels, unless reported otherwise. Note that the more liberal the threshold of an exclusive mask, the more conservative is the masking procedure. For completeness, besides the main results reported below, auxiliary analyses were performed to compare subgroups with different classes of medication, unipolar and bipolar patients type II, or depressed and euthymic patients, that did not shown any modification of the pattern of results (not reported here).

RESULTS

Participants

The two groups did not differ significantly regarding gender, age, laterality and level of education (Table 1). Among patients, 14 were categorized as euthymic, 10 depressed, 3 hypomanic and 5 in sub-depressive/mixed state that did not meet DSM-IV-TR criteria for a Major Depressive Episode. Table 1 summarizes the mean levels of depression, mania and anxiety. Regarding medication, 27 out of 32 patients were taking one or more drugs from one of the class reported in Table 1.

Table 1
Demographic and clinical variables

Behavioral results

Number of words

In total, 14 682 words were produced across the six conditions of interest (close_positive, close_negative, close_neutral, remote_positive, remote_negative and remote_neutral) by HP and patients. Results for each condition and each group are summarized in Table 2. A mixed repeated measures 2 × 3 ANOVA (condition × valence of emotion) with group as between subject factor revealed a main effect of group (F(1,61) = 9.624, P = 0.03) with patients producing fewer words than HP, a main effect of condition (F(1,61) = 8.326, P = 0.005) with fewer words produced for the remote condition, and a main effect of emotion (F(2,60) = 11.348, P < 0.001) with more words produced for the neutral stimuli. There was also a condition by emotion interaction (F(2,60 = 9.534, P < 0.001). No interaction of group × emotion or group × condition was encountered. Post hoc t-tests showed that the close_neutral condition was easier (i.e. resulting in more words) than the other conditions for both groups.

Table 2
Percentage of words produced per condition

Finally, the overall performance correlated negatively with the level of depression (MADRS score, r(63) = −0.67, P < 0.001) and in both groups individually as well (patients: r(32) = −0.657, P < 0.001; HP: r(31) = −0.675, P < 0.001), but not with other clinical questionnaires (e.g. Hamilton Anxiety Scale). As expected, it also correlated positively with level of education (r(63) = 0.275, P = 0.029), but education levels were matched between groups (independent sample t-test: P = 0.365).

Originality

The two indices of originality, FREQ (frequency of the word in the whole dataset) and TYP (typicality compared with norms of associations in the French language), are reported in Table 3 and correlated with each other (close condition: r(63) = 0.906, P < 0.001; remote condition: r(63) = 0.377, P = 0.002). A similar mixed repeated measures (3 emotions × 2 groups) ANOVA was conducted for both indices, for each task condition separately. The frequency of words given per cue (FREQ) for the close condition revealed a main effect of emotion (F(2,60) = 132.72, P < 0.001): participants produced more unusual responses to positive, then negative, then neutral stimuli (P = 0.04 and P < 0.001 for each paired comparison, respectively). Although the group × emotion interaction did not reach significance, there was a tendency for more unusual responses with negative cues in patients compared with controls (P = 0.056, independent sample t-test), but no difference in other emotion conditions.

Table 3
Index of frequency (FREQ) and typicality (TYP) for each condition

For the number of typical associations (TYP) in the close condition, we also found a main effect of emotion (F(2,60) = 43.675, P < 0.001): responses were more unusual for both positive and negative than neutral stimuli. Although the group × emotion interaction did not reach significance, there was a tendency for more unusual responses with negative cues in patients compared with HP (P = 0.062, independent sample t-test, Figure 2). Thus, both indices converged to suggest that patients tended to produce less common/more unique responses than HP for negative cues. Emotion meaning also increased word uniqueness in both groups.

Fig. 2
Behavioral results for indices of originality. (A) close ‘automatic’ condition, TYP index; (B) close ‘automatic’ condition, FREQ index. (C) remote ‘inhibition’ condition, TYP index; (D) remote ‘inhibition’ ...

In the remote condition, the FREQ index also showed a main effect of group: now patients were more usual (F(1,61) = 5.740, P = 0.02, Figure 2), unlike for the close condition. There was again a main effect of emotion (F(2,60) = 8.574, P = 0.001): responses were more unusual for positive, then neutral, then negative stimuli (P = 0.02 and P = 0.043, respectively). For the TYP index in the remote condition, a main effect of group (F(1,61) = 5.55, P = 0.022) also indicated more usual responses in patients than HP in all three emotions (Figure 2).

fMRI results

Main effect of associative tasks

When we compared the two conditions involving semantic-associative processing (close and remote conditions) with the control condition, the whole brain analysis revealed widespread activations within the expected semantic network. Among all participants, activations were found not only in the left middle frontal gyrus, the medial superior frontal cortex, the left operculum, the left superior parietal cortex and the left superior temporal gyrus, but also in the right premotor cortex and the bilateral cerebellum (Figure 3A and Table 4). We then tested for differences between patients and HP, by masking exclusively activations for HP by the same contrast for patients (at P < 0.05). This comparison revealed effects in the left superior frontal cortex, left frontal operculum, bilateral postcentral gyri, as well as supplementary motor area (SMA) and pre-SMA (Table 4), that all were more activated for HP than for patients. The reverse masking procedure yielded no activation.

Fig. 3
Regions activated by conditions of interest. (A) Associative tasks vs control task: left and right frontal gyrus, medial superior frontal gyrus, left superior parietal cortex. (B) Close vs remote conditions: left inferior frontal gyrus, basal ganglia. ...
Table 4
Brain regions showing activations for associative tasks (MNI coordinates; CS: cluster size in number of voxels; HP, healthy participants)

Effect of closeness (close vs remote associations)

Contrasting the close condition with the remote condition (across the two groups), we found activation in the bilateral inferior occipital cortices, left SMA, left inferior frontal gyrus, left middle temporal gyrus and left superior frontal gyrus (Figure 3B, Table 5). The effect of close vs remote associations among HP, masked by the same contrast among patients (P < 0.05), revealed differential activations in left frontal operculum, and also left and right pallidum/putamen, and right caudate nucleus. The latter cortico-subcortical regions were therefore more recruited in HP than patients during the production of more ‘automatic’ associations. The reverse masking procedure yielded no activation.

Table 5
Brain regions showing activations in close and remote conditions (MNI coordinates; CS, cluster size in number of voxels; HP, healthy participants)

Effect of inhibition (remote vs close associations)

The comparison between the remote and the close conditions across groups revealed extensive activations in visual areas including bilateral occipital cortex, the left cuneus and right precuneus, as well as the right orbitofrontal cortex and the right middle/superior frontal gyrus (Table 5). When masking the map for remote > close conditions in HP by the same map for the patients (exclusively at P < 0.05), we found selective increases in the right inferior and middle frontal gyrus, and also in several visual areas (Figure 3C). This suggests that patients were less efficient at recruiting brain regions normally contributing to inhibit automatic responses and generate unusual responses. The reverse masking procedure yielded no activation.

For completeness, we also tested for any correlation of brain activation patterns with the FREQ and TYP indices derived from the participants’ performance, for the close and remote association conditions relative to the control condition. These subsidiary analyses also showed differential recruitment of both occipital and right prefrontal areas in HP and patients (see Supplementary Material), in neighboring but non-overlapping peaks compared with main group differences above.

Effect of emotional stimuli

Finally, in keeping with behavioral effects observed in both controls and patients, we tested whether the emotional content of word cues modulated brain activations during the association task. When contrasting all conditions with emotional stimuli, positive or negative, relative to conditions with neutral stimuli (regardless of the type of association and across both groups), we observed greater activation in widespread regions including visual areas, frontal areas, limbic areas and posterior cingulate cortex (Table 6).

Table 6
Brain regions showing activations for emotional stimuli (MNI coordinates; CS, cluster size in number of voxels; HP, healthy participants)

Unlike other conditions, the group comparison now showed greater responses to emotion in patients relative to controls (i.e. by exclusively masking the activation maps for patients by those for HP, at P < 0.05). These differential increases were predominantly observed in left fusiform gyrus and bilateral parahippocampal gyrus (PHG) (Figure 4 and Table 6), plus left superior and medial frontal gyrus. The inverse group comparison yielded no activation. These effects suggest that the emotional meaning of word cues enhanced the recruitment of regions associated with imagery and memory in patients.

Fig. 4
Effect of emotion vs neutral stimuli. Patients activate more the PHG and medial prefrontal gyrus than HPs for emotional stimuli (all conditions pooled together).

DISCUSSION

Using a free word association task with positive, negative and neutral cues, we showed that people with mood disorders produced fewer words than HP, and that their global performance correlated negatively with the level of depression. Overall, emotional stimuli elicited fewer and more ‘atypical’ answers than neutral stimuli in both groups. However, relative to HP, patients tended to produce less typical (i.e. more original or idiosyncratic) words for the close (automatic) condition, but more typical words for the remote (controlled) condition, with a tendency for this pattern to be stronger for negative material. These findings suggest differences in the retrieval of semantic associations in mood disorder patients, partly modulated by the affective valence of word cues.

Accordingly, at the neural level, our associative task was found to activate a distributed semantic network but these effects were more extensive in HP. Critically, during the close condition, HP activated the basal ganglia (putamen, caudate) and left inferior frontal gyrus, which was not the case for patients. During the remote condition, HP activated the right inferior/middle frontal gyrus, bilateral cuneus/precuneus and visual cortex, again unlike patients. In contrast, only patients activated self-related regions, such as the medial prefrontal cortex (mPFC) and PHG, in response to emotional vs neutral stimuli.

A previous behavioral study using an emotional word association task with depressed college students also reported that more associations were produced in response to neutral words in general, whereas depressed students additionally showed a negative priming bias in word associations (Watkins et al., 1996). These effects might partly be due to the nature of word stimuli. Neutral words more often refer to concrete entities when compared with emotion words, and it is possible that concrete words activate more stereotyped associations, hence the high number and also high typicality of answers to the neutral word cues. Interestingly, here we also found that people with mood disorders tend to produce more unusual words in the close association condition, both for positive and negative cues (a pattern that has been debated; see Isen et al., 1985). We were not able to directly examine the effect of current mood on the pattern of association since our subgroup of patients with hypomania was too small, but the total number of words produced did correlate negatively with depression scores across the whole sample (i.e., in both patients and controls), which is consistent with the idea that positive affect is associated with improved word production (fluency, flexibility) and broader conceptual processing (Fredrickson, 2001).

The reverse pattern of ‘originality’ found between patients and HP for close and remote associations, respectively, fits well with the idea that when the task is more ‘automatic’, the patients might tend to retrieve more idiosyncratic, self- or memory-related information when compared with HP. Thus, the observed increase in atypical associations in the close condition might reflect a facilitation of access to self-related thoughts. This would be consistent with the proposal that associative thinking biases in depression are characterized by greater self focus (Watson et al., 2012), and not just narrower contextual associations (Bar, 2009). In contrast, when the task requires an inhibition of spontaneous associative processes, patients appear more stereotyped. Thus, in the remote condition, they produced more ‘incorrect’ answers (persistence of typical associations), which reflect a relative inability to inhibit the automatic/familiar associations. Again, this may arise either because patients are influenced by personal concerns that were cued by the emotional word meaning and interfered with the task, or because they have less powerful cognitive control. These findings are also consistent with the notion that producing more remote associations takes more time, which might explain why patients produced fewer words in total (Mednick, 1962).

At the brain level, we found that the two associative conditions activated many cortical regions known to be involved in semantic processing and language (Thompson-Schill et al., 1997; Binder et al., 2009). In addition, the close condition (relative to the remote) further activated the left SMA and left inferior frontal cortex, crucially implicated in semantic retrieval (Binder et al., 2009; Wende et al., 2012). Both associative tasks also activated regions of visual cortex, but especially during the remote condition. This is consistent with a contribution of mental imagery during associative processing (Desseilles et al., 2011), either automatically or strategically used by participants in the remote condition, as well as during performance on semantic fluency tasks (Birn et al., 2010).

Both the left inferior frontal cortex and the basal ganglia were activated significantly less (or even not at all) in patients when compared with HP in the close association condition. Basal ganglia are known to contribute to word generation, and the caudate nucleus is involved in lexical retrieval through a left SMA-dorsal caudate-ventral anterior thalamic loop (Crosson et al., 2003) and in language switching in bilinguals (Abutalebi et al., 2008; Wang et al., 2013b). Moreover, besides common symptoms of motor inhibition and apathy, lesions to the basal ganglia may lead to aphasic disorders predominantly affecting automatic speech (such as days of the week or prayers [Ullman, 2004]). These basal ganglia functions might be consistent with an important role in promoting automatic association patterns, as observed here in the close condition for HP. Our results thus point to some impairment in cortico-subcortical circuits in people with mood disorders that may contribute to their difficulties in generating spontaneous verbal associations and possibly new thoughts (Piguet et al., 2010).

In HP, the remote association condition showed specific activations in the right inferior/middle frontal gyrus and right cuneus/precuneus. The former region in right inferior frontal gyrus has consistently been implicated in inhibition processes (Konishi et al., 1999; Depue, 2012). Cuneus/precuneus activations have also been revealed by previous neuroimaging study of Hayling task (Collette et al., 2001; Allen et al., 2008) or verbal suppression (de Zubicaray et al., 2000). These effects therefore support the idea that the remote condition involved an active inhibition of spontaneously generated associations. Moreover, an increase in the cuneus accords with a previous study in bipolar patients (Haldane et al., 2008) that reported that impaired inhibition in Hayling task was linked to reduced cuneus volume. Hence, the lower ability of patients to produce remote associations is associated with lower activation of brain networks involved in response inhibition and visual imagery, two processes crucial to perform the task. These findings suggest deficits in executive cognitive control in mood disorder patients, underpinned by abnormal recruitment of these networks.

In sum, an underactivation of semantic and language production networks in patients appears to underpin their reduced production of words in the close condition (which were also more unusual), whereas underactivation of right prefrontal cortex and medial occipito-parietal areas might account for their difficulty in the remote condition, suggesting that mood disorders involve deficits in both the generation and inhibition of association with distinct neural substrates.

Emotional word processing

Patients recruited additional limbic and paralimbic regions in emotional conditions, including peri-hippocampal and midline cortical structures. A recent study (Laeger et al., 2012) investigating brain response to emotional words in HPs showed greater amygdala activation for both negative and positive words than neutral words, and also a significant correlation for negative words with levels of (subclinical) depression and anxiety. However, this study also revealed greater activation of hippocampus for emotional words (Laeger et al., 2012). This is consistent with previous literature on free word retrieval process (McIntosh et al., 2008; Whitney et al., 2009) and contextual associations (Bar et al., 2007), suggesting modulations by the affective significance of stimuli. In addition, the PHG and mPFC are not only implicated in associative-memory-related processes, but also parts of the default-mode network that has been linked to self-referential processing (D’Argembeau et al., 2005). Increased activations of self-related regions have consistently been replicated in depression (Grimm et al., 2009; Lemogne et al., 2012) and may reflect greater self-focus and negative affective biases in these patients, possibly contributing to rumination and task interference (Marchetti et al., 2012). We speculate that hyperactivation within these regions might interfere with normal thought pattern and spontaneous associative processes, by preventing the retrieval of typical semantic associations in the close condition, and also by disrupting the ability to generate broader associations in the remote condition. This proposal accords with the theoretical hypothesis of Bar (2009) that mPFC hyperactivation in depression might interfere with (or rather bias) the activation in medial temporal structures linked to associative processing.

Limitations

Because our patient population was heterogeneous and included patients with different mood disturbances, we could not determine distinct patterns of changes as a function of different diagnoses or states. However, this choice was motivated by the fact that thought disorders and underlying anomalies in cognitive processes might represent a dimensional impairment in mood disorders across distinct diagnostic categories (Piguet et al., 2010). Moreover, our subsidiary comparison between sub-groups (MDD vs BD, depressed vs euthymics participants) did not elicit any distinctive activations in brain regions that were identified by our main contrasts between conditions. The findings reported here are therefore independent of clinical subgroup or current mood state. A second potential limitation is the difficulty of the ‘remote’ condition that may have limited the opportunity for ‘original’ responses in patients, since it has been suggested that spontaneous associations are facilitated by low cortical arousal states allowing defocused attention (see Martindale and Hasenfus, 1978). Nevertheless, this condition allowed us to test for the differential ability of patients to resist interference by more automatic associative processing, relative to controls and to the close condition. Third, patients generally had more difficulties producing two words in 5 s when compared with HP, leading to a loss of trials with recorded responses and a potential reduction in the statistical power of our analyses. However, this is unlikely to be a confound, since some comparisons (e.g. with emotional word cues) actually revealed stronger effects in patients. Finally, it would certainly be fruitful to carry out a finer analysis of the verbal answers produced by patients; for example, looking at the semantic distance and relationship between words within individuals and within groups.

Conclusion

In mood disorder patients, pattern of free verbal associations might be influenced by a tendency to retrieve more self-related material, particularly in response to emotional cues, making their production less typical in automatic conditions and more typical when answers are controlled. These effects are paralleled with both hypoactivation of brain areas involved in semantic retrieval and inhibitory executive control, and hyperactivation of regions involved in memory and self-related processing. This combination of impaired generation and impaired control of spontaneous associations might contribute to thought disorders in patients (Piguet et al., 2010; Desseilles et al., 2012).

SUPPLEMENTARY DATA

Supplementary data are available at SCAN online.

Conflict of Interest

None declared.

Supplementary Data:

Acknowledgments

The authors thank Alexis Hervais-Adelman for his help with sparse sampling data. C.P., V.S., Y.C. and P.V. are supported by the Swiss National Science Foundation. P.V. is additionally supported by the FOREMANE fund (Société académique). C.P. and M.D. received a grant from the Vachoux Foundation.

REFERENCES

  • Abutalebi J, Annoni J-M, Zimine I, et al. Language control and lexical competition in bilinguals: an event-related FMRI study. 2008 Cerebral Cortex (New York, N.Y.: 1991), 18, 1496–505. [PubMed]
  • Akiskal HS, Benazzi F. Validating Kraepelin’s two types of depressive mixed states: “depression with flight of ideas” and “excited depression” World Journal of Biological Psychiatry. 2004;5:107–13. [PubMed]
  • Allen P, Mechelli A, Stephan KE, et al. Fronto-temporal interactions during overt verbal initiation and suppression. Journal of Cognitive Neuroscience. 2008;20:1656–69. [PubMed]
  • Bar M. A cognitive neuroscience hypothesis of mood and depression. Trends in Cognitive Science. 2009;13:456–63. [PMC free article] [PubMed]
  • Bar M, Aminoff E, Mason M, Fenske M. The units of thought. Hippocampus. 2007;17:420–8. [PubMed]
  • Binder JR, Desai RH, Graves WW, Conant LL. Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies. 2009 Cerebral Cortex (New York, N.Y.: 1991), 19, 2767–96. [PMC free article] [PubMed]
  • Birn RM, Kenworthy L, Case L, et al. Neural systems supporting lexical search guided by letter and semantic category cues: a self-paced overt response fMRI study of verbal fluency. NeuroImage. 2010;49:1099–107. [PMC free article] [PubMed]
  • Burgess PW, Shallice T. Response suppression, initiation and strategy use following frontal lobe lesions. Neuropsychologia. 1996;34:263–72. [PubMed]
  • Collette F, Van der Linden M, Delfiore G, Degueldre C, Luxen A, Salmon E. The functional anatomy of inhibition processes investigated with the Hayling task. NeuroImage. 2001;14:258–67. [PubMed]
  • Crosson B, Benefield H, Cato MA, et al. Left and right basal ganglia and frontal activity during language generation: contributions to lexical, semantic, and phonological processes. Journal of the International Neuropsychological Society. 2003;9:1061–77. [PubMed]
  • D’Argembeau A, Collette F, Van der Linden M, et al. Self-referential reflective activity and its relationship with rest: a PET study. NeuroImage. 2005;25:616–24. [PubMed]
  • De Zubicaray GI, Zelaya FO, Andrew C, Williams SC, Bullmore ET. Cerebral regions associated with verbal response initiation, suppression and strategy use. Neuropsychologia. 2000;38:1292–304. [PubMed]
  • Depue BE. A neuroanatomical model of prefrontal inhibitory modulation of memory retrieval. Neuroscience and Biobehavioral Reviews. 2012;36:1382–99. [PMC free article] [PubMed]
  • Desseilles M, Chang T, Piguet C, Bertschy G, Dayer AG. A three-dimensional model of thoughts: insight into depression. Psychopathology. 2012;45:203–14. [PubMed]
  • Desseilles M, Dang-Vu TT, Sterpenich V, Schwartz S. Cognitive and emotional processes during dreaming: a neuroimaging view. Consciousness and Cognition. 2011;20:998–1008. [PubMed]
  • Drysdale E, Knight HM, McIntosh AM, Blackwood DHR. Cognitive endophenotypes in a family with bipolar disorder with a risk locus on chromosome 4. Bipolar Disorders. 2013;15:215–22. [PubMed]
  • Favre S, Aubry J-M, Gex-Fabry M, Ragama-Pardos E, McQuillan A, Bertschy G. [Translation and validation of a French version of the Young Mania Rating Scale (YMRS)] L’Encéphale. 2003;29:499–505. [PubMed]
  • Ferrand L. Normes d’associations verbales pour 260 mots “abstraits” L’Année Psychologique. 2001;101:683–721.
  • Ferrand L, Alario F-X. Normes d’associations verbales pour 366 noms d’objets concrets. L’Année Psychologique. 1998;98:659–709.
  • First MB, Gibbon M, Spitzer RL, Williams JBW, Benjamin LS. Structured Clinical Interview for DSM-IV Axis II Personality Disorders (SCID-II) Washington, DC: American Psychiatric Press, Inc; 1997.
  • Fredrickson BL. The role of positive emotions in positive psychology. The American Psychologist. 2001;56:218–26. [PMC free article] [PubMed]
  • Gohier B, Ferracci L, Surguladze SA, et al. Cognitive inhibition and working memory in unipolar depression. Journal of Affective Disorders. 2009;116:100–5. [PubMed]
  • Goodwin KF, Jamison RK. Manic-Depressive Illness: Bipolar Disorders and Recurrent Depression. 2nd. New York: Oxford University Press; 2007.
  • Grimm S, Ernst J, Boesiger P, et al. Increased self-focus in major depressive disorder is related to neural abnormalities in subcortical-cortical midline structures. Human Brain Mapping. 2009;30:2617–27. [PubMed]
  • Haldane M, Cunningham G, Androutsos C, Frangou S. Structural brain correlates of response inhibition in Bipolar Disorder I. Journal of Psychopharmacology (Oxford, England) 2008;22:138–43. [PubMed]
  • Hall DA, Haggard MP, Akeroyd MA, et al. “Sparse” temporal sampling in auditory fMRI. Human Brain Mapping. 1999;7:213–23. [PubMed]
  • Hamilton M. The assessment of anxiety states by rating. The British Journal of Medical Psychology. 1959;32:50–55. [PubMed]
  • Isen AM, Johnson MM, Mertz E, Robinson GF. The influence of positive affect on the unusualness of word associations. Journal of Personality and Social Psychology. 1985;48:1413–26. [PubMed]
  • Keizer I, Piguet C, Favre S, et al. Subjective experience of thought overactivation in mood disorders: beyond racing and crowded thoughts. Psychopathology. 2013;47:174–84. [PubMed]
  • Kircher T, Whitney C, Krings T, Huber W, Weis S. Hippocampal dysfunction during free word association in male patients with schizophrenia. Schizophrenia Research. 2008;101:242–55. [PubMed]
  • Konishi S, Nakajima K, Uchida I, Kikyo H, Kameyama M, Miyashita Y. Common inhibitory mechanism in human inferior prefrontal cortex revealed by event-related functional MRI. Brain. 1999;122(Pt 5):981–91. [PubMed]
  • Laeger I, Dobel C, Dannlowski U, et al. Amygdala responsiveness to emotional words is modulated by subclinical anxiety and depression. Behavioural Brain Research. 2012;233:508–16. [PubMed]
  • Lemogne C, Delaveau P, Freton M, Guionnet S, Fossati P. Medial prefrontal cortex and the self in major depression. Journal of Affective Disorder. 2012;136:e1–e11. [PubMed]
  • Manschreck TC, Maher BA, Milavetz JJ, Ames D, Weisstein CC, Schneyer ML. Semantic priming in thought disordered schizophrenic patients. Schizophrenia Research. 1988;1:61–6. [PubMed]
  • Marchetti I, Koster EHW, Sonuga-Barke EJ, De Raedt R. The default mode network and recurrent depression: a neurobiological model of cognitive risk factors. Neuropsychology Review. 2012;22:229–51. [PubMed]
  • Martindale C, Hasenfus N. EEG differences as a function of creativity, stage of the creative process, and effort to be original. Biological Psychology. 1978;6:157–67. [PubMed]
  • Marvel CL, Paradiso S. Cognitive and neurological impairment in mood disorders. The Psychiatric Clinics of North America. 2004;27:19–viii. [PMC free article] [PubMed]
  • Mason MF, Bar M. The effect of mental progression on mood. Journal of Experimental Psychology: General. 2012;141:217–21. [PMC free article] [PubMed]
  • McIntosh AM, Whalley HC, McKirdy J, et al. Prefrontal function and activation in bipolar disorder and schizophrenia. The American Journal of Psychiatry. 2008;165:378–84. [PubMed]
  • Mednick S. The associative basis of the creative process. Psychological Review. 1962;69:220–32. [PubMed]
  • Montgomery SA, Asberg M. A new depression scale designed to be sensitive to change. The British Journal of Psychiatry: The Journal of Mental Science. 1979;134:382–9. [PubMed]
  • Narrow WE, Kuhl EA. Dimensional approaches to psychiatric diagnosis in DSM-5. The Journal of Mental Health Policy and Economics. 2011;14:197–200. [PubMed]
  • Noël X, Van der Linden M, Brevers D, et al. Separating intentional inhibition of prepotent responses and resistance to proactive interference in alcohol-dependent individuals. Drug and Alcohol Dependence. 2013;128:200–5. [PubMed]
  • Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia. 1971;9:97–113. [PubMed]
  • Pellet J, Bobon DP, Mormont I, Lang F, Massardier A. Sous-Echelle Dépression de la CPRS: Masson, Paris; 1980. Etude Princeps de la Validation Française de la MADRS.
  • Pichot P, Pull CB, von Frenckell R, Pull M-C. A factorial analysis of the Hamilton anxiety rating scale. Psychiatrica Fennica. 1981;30:173–183.
  • Piguet C, Dayer A, Kosel M, Desseilles M, Vuilleumier P, Bertschy G. Phenomenology of racing and crowded thoughts in mood disorders: a theoretical reappraisal. Journal of Affective Disorders. 2010;121:189–98. [PubMed]
  • Pronin E, Jacobs E. Thought speed, mood, and the experience of mental motion. Perspectives on Psychological Science. 2008;3:461–485. [PubMed]
  • Schwartz S, Baldo J, Graves RE, Brugger P. Pervasive influence of semantics in letter and category fluency: a multidimensional approach. Brain and Language. 2003;87:400–11. [PubMed]
  • Sheehan DV, Lecrubier Y, Sheehan KH, et al. The mini-international neuropsychiatric interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. The Journal of Clinical Psychiatry. 1998;59(Suppl. 20):22–33. quiz 34–57. [PubMed]
  • Spitzer M. A cognitive neuroscience view of schizophrenic thought disorder. Schizophrenia Bulletin. 1997;23:29–50. [PubMed]
  • Thompson-Schill SL, D’Esposito M, Aguirre GK, Farah MJ. Role of left inferior prefrontal cortex in retrieval of semantic knowledge: a reevaluation. Proceedings of the National Academy of Sciences of the United States of America. 1997;94:14792–7. [PubMed]
  • Ullman MT. Contributions of memory circuits to language: the declarative/procedural model. Cognition. 2004;92:231–70. [PubMed]
  • Vieta E, Valentí M. Mixed states in DSM-5: implications for clinical care, education, and research. Journal of Affective Disorders. 2013;148:28–36. [PubMed]
  • Volle E, de Lacy Costello A, Coates LM, et al. Dissociation between verbal response initiation and suppression after prefrontal lesions. 2012 Cerebral Cortex (New York, N.Y.: 1991), 22, 2428–40. [PMC free article] [PubMed]
  • Wang K, Song L-L, Cheung EFC, Lui SSY, Shum DHK, Chan RCK. Bipolar disorder and schizophrenia share a similar deficit in semantic inhibition: a meta-analysis based on Hayling Sentence Completion Test performance. Progress in Neuro-Psychopharmacology & Biological Psychiatry. 2013a;46:153–60. [PubMed]
  • Wang X, Wang Y-Y, Jiang T, Wang Y-Z, Wu C-X. Direct evidence of the left caudate’s role in bilingual control: an intra-operative electrical stimulation study. Neurocase. 2013b;19:462–9. [PubMed]
  • Watkins PC, Vache K, Verney SP, Muller S, Mathews A. Unconscious mood-congruent memory bias in depression. Journal of Abnormal Psychology. 1996;105:34–41. [PubMed]
  • Watson LA, Berntsen D, Kuyken W, Watkins ER. The characteristics of involuntary and voluntary autobiographical memories in depressed and never depressed individuals. Consciousness and Cognition. 2012;21:1382–92. [PubMed]
  • Wende KC, Straube B, Stratmann M, Sommer J, Kircher T, Nagels A. Neural correlates of continuous causal word generation. NeuroImage. 2012;62:1399–407. [PubMed]
  • Whalley HC, Sussmann JE, Chakirova G, et al. The neural basis of familial risk and temperamental variation in individuals at high risk of bipolar disorder. Biological Psychiatry. 2011;70:343–9. [PubMed]
  • Whitney C, Weis S, Krings T, Huber W, Grossman M, Kircher T. Task-dependent modulations of prefrontal and hippocampal activity during intrinsic word production. Journal of Cognitive Neuroscience. 2009;21:697–712. [PMC free article] [PubMed]
  • Xue G, Aron AR, Poldrack RA. Common neural substrates for inhibition of spoken and manual responses. 2008 Cerebral Cortex (New York, N.Y.: 1991), 18, 1923–32. [PubMed]
  • Young RC, Biggs JT, Ziegler VE, Meyer DA. A rating scale for mania: reliability, validity and sensitivity. British Journal of Psychiatry. 1978;133:429–35. [PubMed]

Articles from Social Cognitive and Affective Neuroscience are provided here courtesy of Oxford University Press